ORIGINAL ARTICLE

doi:10.1111/j.1558-5646.2010.01032.x

THE BITTERLING–MUSSEL COEVOLUTIONARY RELATIONSHIP IN AREAS OF RECENT AND ANCIENT SYMPATRY

Martin Reichard,1,2,3 Matej Polacik,ˇ 2 Ali Serhan Tarkan,4 Rowena Spence,1 Ozcan¨ Gaygusuz,5 Ertan Ercan,4 Marketa´ Ondrackovˇ a,´ 2,6 and Carl Smith1 1School of Biology, University of St Andrews, St Andrews, Fife KY16 8LB, United Kingdom 2Institute of Vertebrate Biology, Academy of Sciences of the Czech Republic, Kvetnˇ a´ 8, 603 65 Brno, Czech Republic 3E-mail: [email protected] 4Faculty of Fisheries, Mugla˘ University, Turkey 5Faculty of Fisheries, Istanbul University, Turkey 6Institute of Botany and Zoology, Faculty of Science, Masaryk University, Kotla´ rskˇ a´ 2, 611 37 Brno, Czech Republic

Received November 9, 2009 Accepted March 22, 2010

Host–parasite relationships are often characterized by the rapid evolution of parasite adaptations to exploit their host, and counteradaptations in the host to avoid the costs imposed by parasitism. Hence, the current coevolutionary state between a parasite and its hosts is predicted to vary according to the history of sympatry and local abundance of interacting species. We compared a unique reciprocal coevolutionary relationship of a fish, the European bitterling (Rhodeus amarus) and freshwater mussels (Unionidae) between areas of recent (Central Europe) and ancient (Turkey) sympatry. Bitterling parasitize freshwater mussels by laying their eggs in the gills of mussel and, in turn, mussel larvae (glochidia) parasitize the fish. We found that all bitterling from both regions avoided one mussel species. Preferences among other mussel species tended to be related to local mussel abundance rather than duration of sympatry. Individual fish were not consistent in their oviposition choices, precluding the evolution of host-specific lineages. Mussels were demonstrated to have evolved strong defenses to bitterling parasitism in the area of ancient sympatry, but have no such defenses in the large areas of Europe where bitterling are currently invasive. Bitterling avoided glochidia infection irrespective of the duration of sympatry.

KEY WORDS: Coevolutionary arm races, evolutionary lag, gentes, host race, specialization, symbiosis.

Host–parasite relationships are often characterized by the recip- In parasites utilizing several hosts, each host species may rocal coevolutionary “arms race” in which parasites show adap- require a distinct adaptation to facilitate exploitation and each tations to maximize exploitation of their host and hosts evolve may evolve distinct counteradaptations. Consequently, it may be counteradaptations to avoid the costs imposed by parasitism, giv- adaptive for an individual parasite that successfully completes ing rise to a complex system of adaptations and counteradapta- development in a specific host to produce progeny that tend to tions (Dawkins and Krebs 1979; Thompson 1994; Poulin 2000). exploit the same host species, which may lead to the evolution of Host and parasite populations are often spatially structured and host-specific lineages (Gibbs et al. 2000; Sorenson et al. 2003; their relationship may evolve to different states across the range Malausa et al. 2005). However, host specialization is not an in- of host–parasite sympatry, depending on the historical and eco- evitable outcome of host–parasite coevolution, because it may logical contexts of the interaction (Gandon and Michalakis 2002; carry costs associated with locating appropriate specific hosts. Thompson and Cunningham 2002; Benkman et al. 2003). Hence it may also be more adaptive for parasites to remain as

C 2010 The Author(s). Journal compilation C 2010 The Society for the Study of Evolution. 3047 Evolution 64-10: 3047–3056 MARTIN REICHARD ET AL.

generalists and opportunistically exploit several host species niles (Smith et al. 2004). Mussels often contain eggs and embryos (Poulin 2000; Silva-Brandao˜ and Solferini 2007). from multiple ovipositions, with up to 250 bitterling embryos in Recent species expansions offer unique scenarios in which a single mussel (Smith et al. 2004). Bitterling embryos can inflict host–parasite coevolution can be compared between situations of significant fitness costs on mussels (Reichard et al. 2006) through ancient and recent sympatry between parasite and host (Lahti damage to gill epithelium, competition with mussels for oxygen, 2006). For example, the brown-headed cowbird (Molothurus and disruption of water circulation over the gills (Stadnichenko ater), a brood parasite of other birds, experienced a dramatic and Stadnichenko 1980; Smith et al. 2001; Mills and Reynolds range expansion and population increase during the last few cen- 2002). turies. This expansion brought it into contact with many new bird Unionid mussels occupy benthic freshwater habitats and fil- species and populations previously unexposed to brood parasitism ter water through their gills to obtain food and oxygen. Water by cowbirds. It appears that many recent host populations fail to enters the mussel gill through an inhalant siphon leading to the respond to parasitism and are exploited as naive hosts that have not mantle cavity. Water is passed through the gills to their inner sur- yet evolved adaptive responses to brood parasitism (evolutionary face and is channeled via the water tubes to the suprabranchial cav- lag) (Rothstein 1990; Soler and Møller 1990). ity from which it is expelled through the exhalant siphon (Bauer Here, we use a unique study system of reciprocal parasitism and Wachtler¨ 2000). The exhalant siphon forms part of the mantle in the European bitterling (Rhodeus amarus), a freshwater fish tissue and is not physically attached to the suprabranchial cavity. species that is an obligate parasite of unionid mussels, laying Hence there is no physical connection between the gills and the its eggs in the gill cavity of the mussel. The mussels used for exhalant siphon and water circulating inside the gills enters the oviposition have larval stages (termed glochidia) that are obli- mantle cavity before reaching the exhalant siphon. Female union- gate parasites of freshwater fish. Both partners display distinctive ids brood their glochidial larvae in modified sections of the gill, morphological, physiological, and behavioral adaptations for host termed marsupia, from which they discharge mature glochidia into exploitation and counteradaptations against being parasitized by the water column. Glochidia are composed of a tiny (0.10 mm in the other partner (Smith et al. 2004). The occurrence of bitterling Unio spp.) hinged valve that snaps shut on contact with fish tissue, in Central and West Europe is recent (Kozhara et al. 2007; Van typically attaching to their gills or fins (Blazekˇ and Gelnar 2006). Damme et al. 2007) and appears to have coincided with an in- When attached, the glochidia must be encysted by host tissue crease in global temperature at the end of the Little Ice Age (cf. to complete their development and remain attached for several 1850 AD) (Van Damme et al. 2007) when bitterling expanded weeks (Bauer and Wachtler¨ 2000). High levels of infection by from the Pontic region (Bohlen et al. 2006). No such range shift glochidia can be lethal to a fish (Meyers and Millemann 1977). was recorded in unionid mussels, which have been a stable compo- Four species of unionid mussels commonly co-occur with nent of the freshwater fauna of Central and West Europe through- bitterling in Central Europe and all of them may be used for out the Holocene (Kennard and Woodward 1903; Kennard 1924; oviposition; Unio pictorum, Unio tumidus, Anodonta anatina,and Watters 2000). Regardless of the absolute timing of the bitterling Anodonta cygnea. Unio crassus co-occurs with bitterling in the expansion to Central and West Europe, the mussel populations Pontic region where it replaces U. tumidus. Host mussel species used by the bitterling for oviposition in Central Europe are hosts differ in the anatomy of their gills, oxygen consumption, and with a recent sympatry in contrast to bitterling populations from flow rate of water circulating through their gills (Smith et al. the Pontic region, where bitterling and mussels have evolved in 2001; Mills et al. 2005). For example, water tubes (the inter- sympatry for at least 2 million years (Bohlen et al. 2006; Kozhara lamellar spaces in the mussel gill where bitterling eggs reside) et al. 2007; Reichard et al. 2007b; Van Damme et al. 2007). of Anodonta spp. are more complex than those of Unio spp. (Liu During reproduction male bitterling court females and lead et al. 2006). Anodonta cygnea has significantly wider gill septa them to mussels for oviposition. Females inspect mussels and, and consumes more oxygen than the other three species (Smith if they choose to oviposit, insert a long ovipositor into the mus- et al. 2001; Mills et al. 2005). These and other differences among sel exhalant siphon to place their eggs deep inside the mussel mussels may be important for the survival of bitterling eggs and gill cavity. Males fertilize the eggs by releasing sperm into the embryos, because the most important source of their mortality inhalant siphon of the mussel, so that water filtered by the mus- is egg ejection from the mussel gills and suffocation (Mills and sel carries the sperm to the eggs. Preoviposition sperm releases, Reynolds 2002; Kitamura 2005). whereby males ejaculate into the siphon of a mussel before a fe- Host mussels have evolved counteradaptations that enable male spawns, are a common feature of male courtship. One to them to eject developing bitterling eggs and embryos (Smith six large elliptical yolk-rich eggs are laid during each spawning. et al. 2004); mussels rapidly contract their valves and expel The eggs hatch in 36 h and embryos develop inside the mussel for a stream of water that can dislodge bitterling eggs and em- three to six weeks before departing as actively swimming juve- bryos from their gills. Bitterling make sophisticated oviposition

3048 EVOLUTION OCTOBER 2010 BITTERLING-MUSSEL COEVOLUTION

decisions that limit ejections (Smith et al. 2000; Mills and This prediction derives from the unequal costs of parasitism for Reynolds 2002). Ejections typically occur within the first six bitterling and mussels. Bitterling are not the only potential host days of embryo development (Nagata 1985; Mills and Reynolds of glochidia, indeed they are often a minor component of the fish 2002; Kitamura 2005). Immediate ejections (within a few sec- assemblages in which they occur and rarely represent a signifi- onds of oviposition) are also an important source of egg mortality cant pool of potential hosts. In contrast, bitterling only occur in (Reichard et al. 2007a) and strongly correlate with overall ejec- sympatry with mussels, and exposure to glochidia represents a tion rate (Reichard et al. 2007b). Ejection of bitterling eggs by potentially substantial cost. mussels is most frequent at the centre of bitterling diversity in Asia (Reichard et al. 2007a), where more than 40 species of bit- terling are recognized and where the bitterling-mussel association Methods is ancient, with bitterling fossils dated to 16 millions years ago STUDY SITES AND POPULATIONS (Tomoda et al. 1977). Data were collected for populations from the River Ballica In central Europe, bitterling almost entirely avoid glochidial (41◦00N, 29◦25E) and Lake Sapanca (40◦42N, 30◦15E) in infection (Reichard et al. 2006), though diverse outcomes are Turkey, and River Kyjovka (48◦47N, 17◦01E) and Lake Bazina reported from other parts of the bitterling-mussel sympatry. Al- (48◦38N, 16◦56E) in the Czech Republic. Populations were though adult and embryo bitterling (Rhodeus spp.) are often heav- selected based on habitat characteristics and the abundance of ily infected by glochidia in the Caucasian region and the Russian unionid mussels (Table 1), with a low and high mussel density Far East (N. Bogutskaya, pers. comm.), Rhodeus ocellatus from site from each region. Bitterling density was high at all sites, with a Southern and Eastern China appear to partially avoid glochidia of range of bitterling to mussel ratios typical of those seen in previous Anodonta woodiana (Dudgeon and Morton 1984; Fukuhara et al. population surveys (Smith et al. 2000). To estimate egg densities 1986; Reichard et al. 2006). within mussels, and thereby the degree of competition for spawn- In this study, we tested the general hypothesis that the ing sites (egg crowding), 10 A. anatina and 10 U. pictorum mussels European bitterling from a region of ancient sympatry with their were exposed to bitterling spawning at each study site over the mussel hosts (the Pontic region of Turkey) are under stronger period of peak bitterling spawning for 3–4 h and expressed as the selection to avoid mussel counteradaptation compared to bitter- number of eggs laid in a mussel per 1 h of exposure (Table 1). ling populations from regions of more recent sympatry where they encounter naive hosts (Central Europe) (Soler and Møller HOST PREFERENCE AND CONSISTENCY 1990). Consequently, we predicted that bitterling from the region Host preference and consistency of preferences were tested in of ancient sympatry would have evolved a stronger preference experimental aquaria (50 × 30 × 30 cm). Each morning, four for particular mussel species, either at the population level or at haphazardly chosen females in spawning condition were selected the level of individual fish (expressed as individual consistency from a stock of recently collected fish (held in 200 L tanks) and of mussel species choice). In the region of ancient sympatry, we placed individually in a transparent plastic cup floating in the further predicted stronger individual host preferences (i.e., con- experimental aquarium in which a single haphazardly captured sistency) in bitterling from sites with high mussel abundance and male from the same population was housed. Each experimental diversity, and population and individual preference for sympatric aquarium contained four mussels in separate sand-filled plastic mussel species in bitterling from sites that contained only a single cups aligned in a row in a randomly selected position. In Turkey, mussel species. For mussel responses to bitterling oviposition, we the mussel species presented were A. anatina (mean ± SD mussel predicted that mussels would have stronger defenses in the region size = 94 ± 15 mm), A. cygnea (119 ± 17), U. crassus (65 ± of ancient sympatry, either as a direct response to prevent oviposi- 6) and U. pictorum (73 ± 7). In the Czech Republic, A. anatina tion or through responses that would ameliorate the cost of hosting (81 ± 12 mm), A. cygnea (118 ± 20), and U. pictorum (77 ± eggs. For infection of bitterling by mussel glochidia, however, we 9) were used along with U. tumidus (73 ± 13), which replaces predicted no difference in the rate of infection between regions. U. crassus in Central and West Europe. Once the male started to

Table 1. Characteristics of four study populations and their collection sites. Mussel species present at sites are ranked from highest to lowest abundance; rarely collected species are in parentheses.

Site L. Sapanca R. Ballica L. Bazina R. Kyjovka

Sympatry Ancient Ancient Recent Recent Mussel species AA, UP, (AC) AA AA, (AC) AA, UP, UT Egg crowding (mean±SE) Low (0.18±0.10) High (4.69±0.80) High (1.02±0.38) Low (0.31±0.11)

EVOLUTION OCTOBER 2010 3049 MARTIN REICHARD ET AL.

court the female she was gently released from the cup and obser- stimulus was measured as the time between exhalant siphon stim- vations started. We collected data on fish behavior toward each ulation and siphon closure. Mussels were placed in sand-filled mussel until oviposition (or for a maximum of 40 min, if oviposi- pots and allowed to settle in a large aquarium with a 200-mm wa- tion did not occur) using established protocols (Smith et al. 2001; ter depth. A metal rod with a blunt tip (diameter 3 mm) was used Reichard et al. 2004). We collected data on male inspection of the to mimic the contact of a bitterling ovipositor during spawning. mussel siphon, male leading of the female to a specific mussel, The response time (to the nearest 0.1 sec) was estimated using sperm release into the mussel inhalant siphon, female inspection footage from a digital video camera (Olympus μ Tough Olym- of the mussel siphon, and female skimming (deceptive oviposi- pus, Tokyo, Japan) replayed, frame-by-frame, using Quick Time tion without releasing eggs). These behaviors represent male and Player. Data on aperture length and tactile reaction could only female investment in specific mussels (Reichard et al. 2007a). If have been measured for mussels from high mussel density sites oviposition occurred, a mussel that received eggs was observed within each region; at low mussel density sites too few mussels for another 15 sec to record any immediate ejections of eggs were collected to obtain adequate sample sizes. (Reichard et al. 2007a, b). Four replicates with four pairs (each in a separate tank) were completed in each round of testing. After GLOCHIDIA INFECTION completion of an observation, mussels were removed, measured Experimental exposure to glochidia was conducted in 40 L for shell length and checked for the presence of glochidia and aquaria, each containing locally collected U. pictorum. The pe- bitterling eggs using a mussel-opening device (Kitamura 2005). riod of glochidia release by U. pictorum, unlike that of A. anatina, A new set of mussels, arranged in a predetermined random order, broadly overlaps with the bitterling reproductive season when the was placed in the tank. Females were captured, transferred into fish are most susceptible to infection as a result of their mussel- another tank in a plastic cup, and tested again with a new male oriented behavior. Each aquarium was stocked with two individ- and new set of mussels. A minimum time between two tests with ual bitterling, a control local cyprinid from another subfamily the same female was 20 min to ensure that females were always (Leuciscinae), and tubenose goby (Proteorhinus marmoratus), ready to oviposit (Smith et al. 2004). A total of 16 replicates with a species known to be highly susceptible to glochidia infection four males and four females were completed, with four replicates (KoubkovaandBaru´ sˇ 2000). Different control cyprinid species completed for each of the test fish (always with a different part- were selected to represent a locally abundant cyprinid, the Dnieper ner). A total of 16 males and 16 females (64 observations) were chub (Petroleuciscus borysthenicus) in Turkey and rudd (Scar- tested from each bitterling population over the period of spawn- dinius erythrophthalmus) in the Czech Republic. Four U. pictorum ing (mid April to mid May). Additional oviposition choice tests with mature glochidia in their gills were added to each of 10 exper- were conducted for the Lake Sapanca population, which showed imental aquaria and left to discharge glochidia. After four days, a low oviposition rate during the initial round of the experiment. mussels were removed and 24 h later one individual of each fish A different set of fish and mussels were used, though the tank species was removed, killed with an overdose of anesthetic and setting was identical. fixed in a 6% formaldehyde solution. The remaining fish were removed and fixed after a further three days. Following physical MUSSEL RESPONSES attachment of glochidia to a fish host, encystation by host tissue During exposure of mussels to natural spawnings to estimate occurs within three days or the fish may eliminate the parasite egg densities in mussels among sites (Table 1), the number and (Wachtler¨ et al. 2000). Hence, our second collection of fish rep- proportion of bitterling eggs deposited in mussel mantle cavi- resented encysted glochidia. An additional seven fish per species ties (instead of their normal positioning in the water tubes or and region were fixed immediately after fish collection to pro- suprabranchial cavity, which is essential for completion of em- vide an estimate of natural infection rates prior to experimental bryo development) was recorded. The aperture length (distance exposure. Only four control P. marmoratus were examined in the between posterior end of the exhalant siphon and the anterior end Czech Republic due to our failure to catch an adequate number. of suprabranchial cavity) was measured to the nearest 0.01 mm Infection rates prior to experimental exposure were negligible (a nondestructively using a mussel-opening device and electronic total of one glochidium in Turkey, three glochidia in the Czech calipers. The aperture is a gap between the gill and mantle tissue Republic). Glochidia were counted on the fish body, fins, gills, through which the bitterling ovipositor passes during oviposition. and gill cover under a dissecting microscope. If the ovipositor enters the aperture, the eggs are laid inside the Statistical analyses were conducted using R (R Core Devel- mantle cavity instead of into the gills. Hence a longer aperture opment Team 2007). Generalized linear models were used to test is predicted to result in more misplaced ovipositions. A total of oviposition preference (binomial error structure), behavioral pref- 25 mussels each of A. anatina and U. pictorum were measured erence (quasi-Poisson error structure), and glochidia load (quasi- from both regions. In addition, the response of mussels to a tactile Poisson error structure). Differences in glochidia load among fish

3050 EVOLUTION OCTOBER 2010 BITTERLING-MUSSEL COEVOLUTION

species were tested for each region separately, different control

Turkey Czech Republic High musseldensity Low mussel density cyprinid species having been used in each. In mussel preference tests, only the first behavior record for each fish was used, to avoid 0.4 Ballica 0.4 Bazina pseudoreplication (the sum of behaviors across all four observa- 0.3 0.3 tions gave concordant results). General linear models were used 0.2 0.2 to test differences in mussel defense mechanisms and consistency 0.1 0.1 in behavioral preference tests. To analyze behavioral consistency, 0 0 Kendall concordance coefficients (W) were calculated for each 0.7 individual fish (for each behavior separately) and subjected to Sapanca 0.4 Kyjovka analysis of variance (ANOVA) after arcsine transformation. Bi- 0.5 0.3

nomial tests and contingency tables were used to test oviposition Proportion of oviposition preference and its consistency. For preferences, random host use 0.3 0.2 served as the null hypothesis (probability of host use 0.25 for each 0.1 0.1 species). Based on mussel preference results, consistency was sub- 0 sequently tested using three host species (A. cygnea was avoided AA AC UC UP AA AC UT UP by bitterling in both regions). Hence, for statistically consistent Figure 1. Host preference of bitterling from four study popula- oviposition choice, the probability of using the same host species tions in terms of the proportion of eggs laid in each of study in sequential choices should have been significantly higher than mussel species (AA, A. anatina;AC,A. cygnea;UC,U. crassus;UP, 0.33, whereas strict consistency would yield a probability of the U. pictorum;UT,U. tumidus). Preferences of all populations are sig- same host use equal to 1. Egg ejection rate was compared between nificant if all four species are considered in the analysis. However, only the Kyjovka population showed a significant preference for regions using binomial tests. U. pictorum and U. tumidus over A. anatina if only three mussel species are considered (details in text). Results HOST PREFERENCE sperm releases and skimming behavior were recorded for this Bitterling significantly avoided A. cygnea for oviposition, but used population. all other host species readily (mussel effect: deviance = 78.95, df = 3, P < 0.001; Fig. 1). There was no effect of mussel size INDIVIDUAL CONSISTENCY IN HOST PREFERENCE (deviance = 2.30, df = 1, P = 0.13) or presence of glochidia (de- There was no consistency in the mussel preferences of ei- viance = 0.80, df = 1, P = 0.37) on bitterling oviposition choice, ther males or females from any bitterling population (Table 3). but an interaction between mussel species and population showed There was also no difference among populations in the consis- that preference for particular mussel species differed among tency of behavioral preferences (Table 2B). The strongest be- bitterling populations (deviance = 17.28, df = 9, P = 0.04). After havioral consistency was recorded for male inspection of mus- exclusion of A. cygnea from the dataset, preferences of bitterling sel siphons (Friedman test, χ2 = 49.38, df = 3, P < 0.001); from sites with high mussel densities were stronger (Kyjovka: other measures of behavioral preference were not consistent deviance = 7.18, df = 2, P = 0.028; Sapanca: deviance = 5.73, (Fig. 2). df = 2, P = 0.057) than from low mussel density sites (Bazina: deviance = 2.30, df = 2, P = 0.318; Ballica: deviance = 2.94, HOST RESPONSES df = 2, P = 0.230). Kyjovka bitterling avoided oviposition in In U. pictorum, the rate of immediate egg ejection was 2.5 times A. anatina, whereas Sapanca bitterling rarely chose U. crassus higher at sites of ancient than recent sympatry (44.4% compared (Fig. 1). to 17.8%), and this difference was significant (binomial test, χ2 = Behavioral preferences were generally congruent with 4.73, df = 1, P = 0.030), although no difference was observed in oviposition choice. Details of statistical tests (GLLM with A. anatina (binomial test, χ2 = 0.001, df = 1, P = 0.976; 21.8% quasi-Poisson distribution) are summarized in Table 2A. Overall, overall). fish paid significantly less attention to A. cygnea. Significant in- Significantly more eggs were found in the mantle cavity, out- teractions between site and mussel species revealed that fish from side the mussel gill, at sites of ancient sympatry in both A. anatina the Kyjovka inspected U. tumidus most often compared to a high (binomial test, χ2 = 129.6, d.f. = 1, P < 0.001) and U. pictorum inspection rate of A. anatina by females from populations in Bal- (binomial test, χ2 = 160.3, d.f. = 1, P < 0.001). This phenomenon lica and Bazina (Table 2). Although fish from Sapanca inspected was observed in 63% of A. anatina and 53% of U. pictorum, affect- mussels readily and males led females to them, significantly fewer ing 34% of recently laid eggs in A. anatina and 22% of recently

EVOLUTION OCTOBER 2010 3051 MARTIN REICHARD ET AL.

Table 2. Results of (A) the effect of population identity, mussel species, and their interaction on behavioral preference measures in four bitterling populations using GLLM analysis with quasi-Poisson distribution. Presence of glochidia in mussel gills, mussel shell size, and presence of bitterling eggs were included in the analysis as uncontrolled sources of variation. (B) The effect of the length of the bitterling-mussel association and mussel density on behavioral consistency (ANOVA on Kendall coefficients of concordance). Smaller degrees of freedom for leading and skimming behavior come from a failure of some fish to perform these behaviors.

(A) Behavioral preference (B) Behavioral consistency

df FP df FP

Leading Locality 3,235 2.34 0.074 Sympatry 1,59 1.24 0.271 Mussel species 3,235 7.90 <0.001 Density 1,59 0.16 0.688 Glochidia 1,235 0.01 0.924 Symp∗dens 1,59 1.64 0.205 Mussel size 1,235 1.36 0.245 Egg presence 1,235 1.07 0.301 Locality × mussel sp 9,235 0.90 0.530 Sperm release Locality 3,235 3.03 0.030 Sympatry 1,60 1.20 0.277 Mussel species 3,235 10.39 0.001 Density 1,60 0.16 0.686 Glochidia 1,235 0.29 0.589 Symp∗dens 1,60 0.91 0.344 Mussel size 1,235 0.01 0.912 Egg presence 1,235 0.22 0.635 Locality × mussel sp 9,235 1.55 0.132 Male inspection Locality 3,235 4.47 0.005 Sympatry 1,60 0.14 0.706 Mussel species 3,235 27.36 <0.001 Density 1,60 1.34 0.251 Glochidia 1,235 1.08 0.299 Symp∗dens 1,60 0.33 0.570 Mussel size 1,235 1.06 0.305 Egg presence 1,235 0.002 0.964 Locality × mussel sp 9,235 2.84 0.003 Female inspection Locality 3,235 2.35 0.073 Sympatry 1,59 0.25 0.619 Mussel species 3,235 6.33 <0.001 Density 1,59 1.93 0.170 Glochidia 1,235 1.64 0.201 Symp∗dens 1,59 1.57 0.215 Mussel size 1,235 0.05 0.826 Egg presence 1,235 0.19 0.667 Locality × mussel sp 9,235 2.70 0.005 Skimming Locality 3,235 6.55 <0.001 Sympatry 1,48 0.07 0.790 Mussel species 3,235 3.27 0.022 Density 1,48 0.83 0.367 Glochidia 1,235 0.07 0.792 Symp∗dens 1,48 0.25 0.618 Mussel size 1,235 0.79 0.374 Egg presence 1,235 5.76 0.017 Locality × mussel sp 9,235 1.51 0.146

laid eggs in U. pictorum (Fig. 3). In four A. anatina, all the eggs 0.093), and no interaction between species and region (F1,42 = were positioned in the mantle cavity. In the area of recent sym- 0.21, P = 0.650). patry, eggs were rarely positioned in the mantle cavity; only two Mussels at sites of ancient sympatry had longer apertures eggs in a single A. anatina and a single egg in U. pictorum. (two-way analysis of covariance [ANCOVA], with mussel size as

Mussels from sites of ancient sympatry responded to a tactile a covariate, F1,95 = 23.02, P < 0.001) and U. pictorum had longer stimulus with a more rapid closure of their exhalant siphons (two- apertures than A. anatina (F1,95 = 36.65, P < 0.001). There was way ANOVA, F1,42 = 18.62, P < 0.001), although there was no no significant interaction between these variables (F1,95 = 0.64, difference between A. anatina and U. pictorum (F1,42 = 2.96, P = P = 0.427).

3052 EVOLUTION OCTOBER 2010 BITTERLING-MUSSEL COEVOLUTION

Table 3. Consistency of bitterling oviposition behavior measured as repeated use of the same mussel species in two consecutive ovipositions with different individual mussels (cons). Given that A. cygnea was avoided in all populations, consistency was defined as the probability of using the same host species at a significantly greater probability than 0.33.

Females Males Cons Total Prop χ2 P Cons Total Prop χ2 P

Ballica 13 40 0.33 0.008 0.927 15 40 0.38 0.208 0.648 Sapanca 1 4 0.25 – – – – – – – Bazina 11 32 0.34 0.010 0.919 10 33 0.30 0.091 0.763 Kyjovka 9 29 0.31 0.046 0.830 8 28 0.29 0.190 0.663

GLOCHIDIA INFECTION in prevalence was observed between sampling periods. Prevalence Glochidial infection of bitterling was low compared to control in the region of ancient sympatry was 90% in P. marmoratus, species in both regions (GLLM with quasi-Poisson distribution, 75% in the control cyprinid, and 45% in bitterling, and in recent

F2,57 = 9.73, P < 0.001 and F2,56 = 17.01, P < 0.001 for sites of sympatry 84% in P. marmoratus, 80% in control cyprinid, and ancient and recent sympatry, respectively). There was no effect of 25% in bitterling. host size on the number of glochidia on a fish (F2,55 = 1.91, P = = = 0.173 and F1,54 0.27, P 0.608; ancient and recent sympatry, Discussion respectively). There was also no significant difference between We compared the reciprocal coevolutionary relationship between the glochidia counts 24 or 72 h after infection from the site of bitterling fish and freshwater mussels in areas of ancient and ancient sympatry (F , < 0.01, P = 0.979), although there was a 1 56 recent sympatry. Our key findings were: (1) all bitterling popula- significant decrease in the site of recent sympatry (F , = 5.29, 1 56 tions from both regions avoided one species of host mussel, but P = 0.026; Fig. 4). used the other three mussel species readily; (2) mussel defense The prevalence of glochidia was similarly low in bitterling was stronger in a region of ancient sympatry than of recent sym- compared to control species and not significantly different be- patry; (3) bitterling avoided glochidia infection irrespective of the tween sites of ancient and recent sympatry, although no decrease duration of sympatry. We found that all bitterling populations avoided oviposition 0.7 in A. cygnea, which is consistent with previous studies from West- ern and Central Europe (Reynolds et al. 1997; Smith et al. 2000). 0.6 Anodonta cygnea has the widest gill septa, consumes most oxy- gen, and has the highest speed of water circulation through its gills than the other three species (Smith et al. 2001; Mills and Reynolds 0.5 2002; Mills et al. 2005), with the result that the mortality rate of bitterling embryos in this mussel is significantly higher than the 0.4 A B 0.3 45 proportion of mussels 75 proportion of eggs with eggs in mantle cavity in mantle cavity

Kendall coefficient of concordance Kendall 30 0.2 50 Leading Inspection Skimming Sperm release Inspection 15 25 MALE FEMALE 0 0 Figure 2. Individual consistency of particular behaviors directed A. anatina U. pictorum A. anatina U. pictorum toward specific mussels across four observations of each individ- TUR CZ TUR CZ TUR CZ TUR CZ ual fish (with different individual mussel and a different mating partner in each observation) measured as Kendall coefficient of Figure 3. The proportion of (A) mussels with at least a single egg concordance (mean ± 1 standard error (box) and confidence in- in its mantle cavity, and (B) the overall proportion of eggs in the terval (whiskers)). Male behavior is represented by black boxes, mantle cavity (across all mussels pooled) for A. anatina (AA) and female behavior by white boxes. U. pictorum (UP) in areas of ancient (TUR) and recent (CZ) sympatry.

EVOLUTION OCTOBER 2010 3053 MARTIN REICHARD ET AL.

60 (Liu et al. 2006; Reichard et al. 2007a; Kitamura et al. 2009) and frequent occurrence of host-specific lineages in avian brood par- 50 asites (Kruger et al. 2009), plant-feeding insects (Dres` and Mallet

40 2002), and other comparable systems (e.g., Munday et al. 2004). However, this prediction was not satisfied and may simply reflect 30 an overall shorter period of association between bitterling and mussels in Europe, compared with Asia (Tomoda et al. 1977). 20 Alternatively, the benefits associated with consistent host prefer- 10 ences may be exceeded by the costs of specializing on a single or Mean abundance of glochidia limited number of hosts. 0 goby cyprinid bitterling goby cyprinid bitterling Mussel defenses were strikingly stronger in the region of an- Turkey Czech Republic cient sympatry. We previously suggested that bitterling in Central

Figure 4. Glochidia load of bitterling and two control species Europe use relatively naive hosts that have not evolved strong (goby—Proteorhinus marmoratus, cyprinid—Petroleuciscus borys- counteradaptations to defend against parasitization by bitterling thenicus in Turkey and Scardinius erythrophthalmus in the Czech (Reichard et al. 2007b). Based on mussel defenses against bit- Republic) in terms of mean glochidia abundance per dissected fish terling parasitism in Asia (Reichard et al. 2007a), we predicted one day (dashed line) and four days (solid line) after exposure to egg ejection behavior to be the most important adaptation of mus- infection. sels in Turkey. This prediction was met in U. pictorum, with 2.5 times more frequent ejections of eggs following oviposition other species available (Smith et al. 2000, 2001). The finding that in Turkish populations in comparison with Czech populations, host species preferences are congruent across the European bitter- although not in A. anatina. Strikingly, we observed high numbers ling range suggests that there are clear physiological constraints of eggs in the mantle cavity of mussels of both species in Turkey; to host quality outside the coevolutionary relationship, with bitter- a situation rarely observed in European mussels. Accordingly, we ling embryo adaptations appearing suboptimal for development in tested the two most likely mechanisms that could be responsible A. cygnea. Despite their low quality as hosts, bitterling embryos for the presence of bitterling eggs in the mantle cavity. These are able to complete development in A. cygnea andtheyareused experiments demonstrated that in the region of ancient sympa- as hosts, albeit at a low frequency (Reynolds et al. 1997; Smith try, mussels closed their siphons significantly more quickly in et al. 2000). response to tactile stimulus, which would tend to interrupt the Bitterling preferences for the remaining three mussel species correct insertion of the ovipositor into the mussel gill cavity. In varied only slightly among populations. A prediction that bit- addition, mussel internal anatomy differed between regions, with terling from sites with only a single host species would prefer a significantly larger distance between the gill and mantle tissues sympatric mussels was not confirmed, strengthening the conclu- (termed aperture length) in the area of ancient sympatry. We pre- sion that the mussel preferences of the European bitterling are sumed that the larger aperture length more frequently leads to not shaped by learning or imprinting. Imprinting on natal host the misplacement of ovipositor during spawning and may further species is reported to maintain host-specific lineages in African increase incidence of the egg deposition outside the gills. indigobirds (Sorenson et al. 2003), but in other study systems, We assumed that the difference in mussel defense against bit- proximate mechanisms of host discrimination are more difficult terling oviposition represented enhanced mussel resistance in the to test. For example, it is known that female European cuck- coevolutionary interaction. However, our experimental design did oos (Cuculus canorus) consistently lay their eggs into nests of not reveal whether the observed resistance is specific to the ob- host species in which they were raised, leading to the evolu- served sympatric combination or whether it is a broad resistance tion of female host-specific races (called gentes) (Gibbs et al. mechanism against bitterling from any region. To investigate the 2000), although the underlying mechanism is not known de- specificity of mussel response, a cross-infection experiment test- spite a substantial research effort (reviewed in Langmore and ing the response of mussel populations to bitterling oviposition Kilner 2007). between regions of recent and ancient sympatry must be em- Populations did not differ in their consistency of host pref- ployed. Such experiments could further unambiguously test other erence, and individual males and females were inconsistent with hypotheses, for example whether bitterling from regions of re- respect to host species in their behavioral and oviposition prefer- cent sympatry have evolved an enhanced ability to use mussels ences. We predicted a higher level of consistency in populations more effectively than in regions of ancient sympatry (although of ancient sympatry, given the number of bitterling species or interpopulation differences in mussel behavior and anatomy are subspecies that are specialized in their use of host species in Asia not congruent with this view). The prediction for cross-infection

3054 EVOLUTION OCTOBER 2010 BITTERLING-MUSSEL COEVOLUTION

experiments is that bitterling from regions of recent sympatry is that use of mussel species by bitterling may be relatively gener- would face the same resistance from mussels in ancient sympatry alist at the edge of the range of bitterling-mussel sympatry, where as local bitterling if only mussel resistance differs between the re- mussel responses are limited. In contrast, at the centre of their gions. However, if both partners coevolve and bitterling from the association, there may be stronger selection for specialization by region of ancient sympatry have partly compensated for enhanced bitterling where mussel responses are stronger and interspecific mussel defense, bitterling from the region recent sympatry are ex- competition for mussels among bitterling species widespread. pected to be even less successful in their utilization of mussels from the region of ancient sympatry. Bitterling avoided glochidia infection in both regions. The ACKNOWLEDGMENTS The study was supported by the Leverhulme Trust and GA AV ability to avoid glochidia infection could be either acquired or (KJB600930802). Experimental procedures were approved by ethical genetic (or both). Bitterling embryos are incubated in intimate committees of the IVB and GAAVand were in accordance with Czech and contact with mussel tissue throughout their development, al- Turkish legal requirements. We thank the Sapanca Inland Fish Aquacul- though rarely directly with glochidia. Fish develop immunity ture & Research Application Station of the Fisheries Faculty of Istanbul University and the Republic of Turkey Ministry of Agriculture and Rural to glochidia infection and become less susceptible to repeated Affairs for facilities and permits necessary to conduct the study. We are exposure (Jansen et al. 2000; Rogers and Dimock 2003). A sig- grateful to J. Graves, M. Honza, and A. Magurran for helpful discussion nificant proportion of experimental bitterling probably developed and comments on the manuscript. in A. anatina, but all appeared unsusceptible to Unio glochidia in- MR and CS conceived and designed the study and interpreted the results, MR, CS, MP, and RS collected data, MO counted glochidia, AST, fection, suggesting that any potential acquired immunity applies OG, and EE contributed to field and experimental work and facilitated across host mussel genera. Alternatively, there may be a genetic research. MR analyzed the data and drafted the manuscript. component to resistance. Bitterling that expanded from areas of long-term sympatry to exploit naive mussel populations do not LITERATURE CITED appear to have experienced a relaxation in selection pressure to Abrams, P. A., and T. J. Kawecki. 1999. Adaptive host preference and the avoid glochidia infection. This finding suggests that the cost of dynamics of host-parasitoid interactions. Theor. Popul. Biol. 56:307– harboring glochidia would be high enough to generate an evolved 324. genetic response. Whether the response is acquired or genetic can Bauer, G., and K. Wachtler.¨ 2000. Ecology and evolution of the freshwater mussels unionoida. Springer-Verlag, Berlin. perhaps be tested using in vitro spawning and artificial exposure Benkman, C. W., T. L. Parchman, A. Favis, and A. M. Siepielski. 2003. Re- of laboratory-reared bitterling. The mechanism of resistance to ciprocal selection causes a coevolutionary arms race between crossbills glochidia by bitterling is not known. and lodgepole pine. Am. Nat. 162:182–194. In conclusion, we established that mussels have evolved Blazek,ˇ R., and M. Gelnar. 2006. Temporal and spatial distribution of strong defenses to bitterling parasitism in an area of ancient sym- glochidial larval stages of European unionid mussels (Mollusca: Union- idae) on host fishes. Folia Parasitol. 53:98–106. patry, whereas their defenses appear to be more relaxed in an area Bohlen, J., V. Slechtovˇ a,´ N. Bogutskaya, and J. Freyhof. 2006. Across Siberia of relatively recent sympatry. This finding contradicts a previ- and over Europe: phylogenetic relationships of the freshwater fish genus ous conclusion that mussel defense behavior might not have been Rhodeus in Europe and the phylogenetic position of R. sericeus from shaped by coevolutionary interactions with bitterling and is rather the River Amur. Mol. Phylogenet. Evol. 40:856–865. Dawkins, R., and J. R. Krebs. 1979. Arms races between and within species. a byproduct of mussel physiology (Mills et al. 2005). We fur- Proc. R. Soc. Lond. B 205:489–511. ther demonstrated that bitterling adaptation to mussels is likely Dres,` M., and J. Mallet. 2002. Host races in plant-feeding insects and their to be generalized, with no host specificity at the individual or importance in sympatric speciation. Phil. Trans. R. Soc. B 357:471–492. population level. Two explanations may account for this finding. Dudgeon, D., and B. Morton. 1984. Site selection and attachment duration of Anodonta woodiana (Bivalvia: Unionacea) glochidia on fish hosts. J. First, generalized bitterling embryo adaptations (a unique wedged Zool. 204:355–362. shape, presence of scaly tubercles on the yolk, positive rheo- Fukuhara, S., Y. Nagata, and T. Yamada. 1986. Glochidium parasitic period, taxis and negative phototaxis) (Smith et al. 2004) are a general host-fish and parasitic site of Anodonta woodiana in small pond. Japan. response to prevent ejections with similar effectiveness among J. Malacol. 45:43–52. all three preferred host species. Second, there are constraints on Gandon, S., and Y. Michalakis. 2002. Local adaptation, evolutionary potential and host-parasite coevolution: interactions between migration, mutation, host specialization due to temporal changes in the abundance of population size and generation time. J. Evol. Biol. 15:451–462. particular mussel species or migration between bitterling popula- Gibbs, H. L., M. D. Sorenson, K. Marchetti, M. D. Brooke, N. B. Davies, and tions with different host availability (Abrams and Kawecki 1999; H. Nakamura. 2000. Genetic evidence for female host-specific races of Gandon and Michalakis 2002). Notably, there is a considerably the common cuckoo. Nature 407:183–186. Jansen, W., G. Bauer, and E. Zahner-Meike. 2000. Glochidial mortality in greater diversity of mussel and bitterling species in Asia where freshwater mussels. Pp. 185–211 in G. Bauer, and K. Wachtler,¨ eds. several bitterling species have specialized to exploit a single mus- Ecology and evolution of the freshwater mussels unionoida. Springer- sel species (Liu et al. 2006; Reichard et al. 2007a). One possibility Verlag, Berlin.

EVOLUTION OCTOBER 2010 3055 MARTIN REICHARD ET AL.

Kennard, A. S. 1924. The Pleistocene non-marine Mollusca of England. Proc. that bitterling fish (Rhodeus sericeus) are parasites of unionid mussels Malacol. Soc. 16:84–97. in Europe. J. Evol. Biol. 19:788–796. Kennard, A. S., and B. B. Woodward. 1903. On the Mollusca of some Holocene Reichard, M., H. Z. Liu, and C. Smith. 2007a. The co-evolutionary relation- deposits of the Thames River system. Proc. Malacol. Soc. 18:90–96. ship between bitterling fishes and freshwater mussels: insights from Kitamura, J. 2005. Factors affecting seasonal mortality of rosy bitterling interspecific comparisons. Evol. Ecol. Res. 9:239–259. (Rhodeus ocellatus kurumeus) embryos on the gills of their host mussel. Reichard, M., M. Przybylski, P. Kaniewska, H. Liu, and C. Smith. 2007b. A Pop. Ecol. 47:41–51. possible evolutionary lag in the relationship between freshwater mussels Kitamura, J., T. Abe, and J. Nakajima. 2009. The reproductive ecology of two and European bitterling. J. Fish Biol. 70:709–725. subspecies of the bitterling Rhodeus atremius (Cyprinidae, Acheilog- Reynolds, J. D., V. J. Debuse, and D. C. Aldridge. 1997. Host specialisation nathinae). Ichthyol. Res. 56:156–161. in an unusual symbiosis: European bitterlings spawning in freshwater Koubkova,´ B., and V. Barus.ˇ 2000. Metazoan parasites of the recently estab- mussels. Oikos 78:539–545. lished tubenose goby (Proterorhinus marmoratus: Gobiidae) population Rogers, C. L., and R. V. Dimock. 2003. Acquired resistance of bluegill sun- from the South Moravian reservoir, Czech Republic. Helminthologia fish Lepomis macrochirus to glochidia larvae of the freshwater mussel 37:89–95. Utterbackia imbecillis (Bivalvia: Unionidae) after multiple infections. Kozhara, A. V., A. V. Zhulidov, S. Gollasch, M. Przybylski, V. G. Poznyak, J. Parasitol. 89:51–56. D. A. Zhulidov, and T. Y. Gurtovaya. 2007. Range extension and con- Rothstein, S. I. 1990. A model system for coevolution—avian brood para- servation status of the bitterling, Rhodeus sericeus amarus in Russia and sitism. Annu. Rev. Ecol. Syst. 21:481–508. adjacent countries. Folia Zool. 56:97–108. Silva-Brandao,˜ K. L., and V. N. Solferini. 2007. Use of host plants by Troidini Kruger,¨ O., M. D. Sorenson, and N. B. Davies. 2009. Does coevolution butterflies (Papilionidae, Papilioninae): constraints on host shift. Biol. promote species richness in parasitic cuckoos? Proc. R. Soc. Lond. B J. Linn. Soc. 90:247–261. 276:3871–3879. Smith, C., J. D. Reynolds, and W. J. Sutherland. 2000. Adaptive host choice Lahti, D. C. 2006. Persistence of egg recognition in the absence of cuckoo and avoidance of superparasitism in the spawning decisions of bitterling brood parasitism: pattern and mechanism. Evolution 60:157–168. (Rhodeus sericeus). Behav. Ecol. Sociobiol. 48:29–35. Langmore, N. E., and R. M. Kilner. 2007. Breeding site and host selection by Smith, C., K. Rippon, A. Douglas, and P. Jurajda. 2001. A proximate cue for Horsfield’s bronze-cuckoos, Chalcites basalis. Anim. Behav. 74:995– oviposition site choice in the bitterling (Rhodeus sericeus). Freshwat. 1004. Biol. 46:903–911. Liu, H. Z., Y.R. Zhu, C. Smith, and M. Reichard. 2006. Evidence of host speci- Smith, C., M. Reichard, P. Jurajda, and M. Przybylski. 2004. The reproductive ficity and congruence between phylogenies of bitterlings and freshwater ecology of the European bitterling (Rhodeus sericeus). J. Zool. 262:107– mussels. Zool. Stud. 45:428–434. 124. Malausa, T., M.-T. Bethenod, A. Bontemps, D. Bourguet, J.-M. Cornuet, and Soler, M., and A. P. Møller. 1990. Duration of sympatry and coevolution S. Ponsard. 2005. Assortative mating in sympatric host races of the between the great spotted cuckoo and its magpie host. Nature 343:748– European corn borer. Science 308:258–260. 750. Meyers, T. R., and R. E. Millemann. 1977. Glochidiosis of salmonid fishes. 1. Sorenson, M. D., K. M. Sefc, and R. B. Payne. 2003. Speciation by host switch Comparative susceptibility to experimental infection with Margaritifera in brood parasitic indigobirds. Nature 424:928–931. margaritifera (L.) (Pelecypoda-Margaritiferidae). J. Parasitol. 63:728– Stadnichenko, A. P., and Y. A. Stadnichenko. 1980. On the effect of bitter- 733. ling larvae on the lamellibranchid mollusc Unio rostratus gentilis Haas. Mills, S. C., and J. D. Reynolds. 2002. Mussel ventilation rates as approxi- Gidrobiologicheski Zhurnal 1980:57–61. mate cue for host selection by bitterling, Rhodeus sericeus. Oecologia Thompson, J. N. 1994. The coevolutionary process. Univ. of Chicago Press, 131:473–478. Chicago. Mills, S. C., M. I. Taylor, and J. D. Reynolds. 2005. Benefits and costs to Thompson, J. N., and B. C. Cunningham. 2002. Geographic structure and mussels from ejecting bitterling embryos: a test of the evolutionary dynamics of coevolutionary selection. Nature 417:735–738. equilibrium hypothesis. Anim. Behav. 70:31–37. Tomoda, Y., H. Kodera, T. Nakajima, and T. Yasuno. 1977. Fossil freshwater Munday, P. L., L. van Herwerden, and C. L. Dudgeon. 2004. Evidence for fishes from Japan. Tisicigarcuronsju 14:221–243. sympatric speciation by host shift in the sea. Curr. Biol. 14:1498–1504. Van Damme, D., N. Bogutskaya, R. C. Hoffmann, and C. Smith. 2007. The Nagata, Y. 1985. The effect of social organization of the bitterling, Rhodeus introduction of the European bitterling Rhodeus amarus to west and ocellatus (Kner) and the condition of a bivalve, Anodonta woodiana central Europe. Fish Fisher. 8:79–106. Lea in a small pond on the number of fish eggs being laid in a bivalve. Wachtler,¨ K., M. C. Dreher-Mansur, and T. Richter. 2000. Larval types and Memoirs of Osaka Kyoiku Univ. 34:9–26. early postlarval biology of naiads (Unionida). Pp. 93–125 in G. Bauer, Poulin, R. 2000. Evolutionary ecology of parasites. Princeton Univ. Press, and K. Wachtler,¨ eds. Ecology and evolution of the freshwater mussels Princeton, NJ. unionoida. Springer-Verlag, Berlin. R Development Core Team. 2006. R: A language and environment for statis- Watters, G. T. 2000. The evolution of Unionacea in North America, and tical computing. R Foundation for Statistical Computing, Vienna. its implications for the worldwide fauna. Pp. 281–309 in G. Bauer, Reichard, M., P. Jurajda, and C. Smith. 2004. Male-male interference com- and K. Wachtler,¨ eds. Ecology and evolution of the freshwater mussels petition decreases spawning rate in the European bitterling (Rhodeus unionoida. Springer-Verlag, Berlin. sericeus). Behav. Ecol. Sociobiol. 56:34–41. Reichard, M., M. Ondrackovˇ a,´ M. Przybylski, H. Liu, and C. Smith. 2006. The costs and benefits in an unusual symbiosis: experimental evidence Associate Editor: S. Gandon

3056 EVOLUTION OCTOBER 2010